Tuning a dynamic positioning system

Abstract The tuning of a complex control system used in ships and semisubmersible oil drilling rigs was studied. A method was sought which minimizes the time needed to tune a manned ship. Three different methods were compared to estimate the parameters of the nonlinear state filters used in the control system. Simulations showed that the modified extended Kalman filter did not yield correct parameter values, mainly because the coloured noise originating from the sea waves dominated the estimation residuals. Simple methods based on average propeller forces and ship accelerations were tested by simulation and on a real scale on two ships. An expert system was used for tuning. Experiments showed that the wave forces disturb the acceleration tests too much. Finally, the tuning problem was solved by using the steepest descent optimization method to minimize the square sum of the Kalman filter estimation residuals. Nonlinear optimization seems to be the most promising method for tuning the parameters of the dynamic positioning (DP) system used in off-shore operations. The collection of data over a short period in sea trials followed by tuning of the control system off-line greatly reduces the costs of sea trials. The benefits of expert system techniques were not achieved because heavy signals analysis and optimization computations were required. The tasks where expert systems could be used were so simple that an ordinary high level language was used.